<div dir="ltr">Why did you revert default sysroot poisoning?</div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Jan 12, 2015 at 5:18 AM, Lei, Maohui <span dir="ltr"><<a href="mailto:leimaohui@cn.fujitsu.com" target="_blank">leimaohui@cn.fujitsu.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Armin<br>
<span class=""><br>
> I think all the two patches need an "Upstream-Status:" statement.<br>
><br>
<br>
</span>OK. I will modify my patch and submit it again.<br>
<br>
<br>
Cheers<br>
<br>
Lei Maohui<br>
<div><div class="h5"><br>
<br>
<br>
> -----Original Message-----<br>
> From: akuster808 [mailto:<a href="mailto:akuster808@gmail.com">akuster808@gmail.com</a>]<br>
> Sent: Monday, January 12, 2015 12:09 PM<br>
> To: Lei, Maohui/雷 茂慧; <a href="mailto:openembedded-core@lists.openembedded.org">openembedded-core@lists.openembedded.org</a><br>
> Subject: Re: [OE-core] [PATCH] gcc: upgrade to 4.9.2 and fix the bug #6824<br>
><br>
> Lei,<br>
><br>
> I think all the two patches need an "Upstream-Status:" statement.<br>
><br>
> - Armin<br>
><br>
> On 01/11/2015 07:12 PM, Lei Maohui wrote:<br>
> > fix the yocto bug #6824<br>
> > backport from gcc<br>
> ><br>
> > Signed-off-by: Lei Maohui <<a href="mailto:leimaohui@cn.fujitsu.com">leimaohui@cn.fujitsu.com</a>><br>
> > ---<br>
> > meta/recipes-devtools/gcc/gcc-4.9.inc | 12 +-<br>
> > ...e500-double-in-SPE_SIMD_REGNO_P-registers.patch | 13 ++<br>
> > .../0061-Fix-for-unwinder-aborts-on-e500.patch | 161<br>
> +++++++++++++++++++++<br>
> > 3 files changed, 180 insertions(+), 6 deletions(-)<br>
> > create mode 100644<br>
> meta/recipes-devtools/gcc/gcc-4.9/0060-Only-allow-e500-double-in-SPE_SIMD_<br>
> REGNO_P-registers.patch<br>
> > create mode 100644<br>
> > meta/recipes-devtools/gcc/gcc-4.9/0061-Fix-for-unwinder-aborts-on-e500<br>
> > .patch<br>
> ><br>
> > diff --git a/meta/recipes-devtools/gcc/gcc-4.9.inc<br>
> > b/meta/recipes-devtools/gcc/gcc-4.9.inc<br>
> > index 2568e99..c4ec40d 100644<br>
> > --- a/meta/recipes-devtools/gcc/gcc-4.9.inc<br>
> > +++ b/meta/recipes-devtools/gcc/gcc-4.9.inc<br>
> > @@ -2,11 +2,11 @@ require gcc-common.inc<br>
> ><br>
> > # Third digit in PV should be incremented after a minor release<br>
> ><br>
> > -PV = "4.9.1"<br>
> > +PV = "4.9.2"<br>
> ><br>
> > # BINV should be incremented to a revision after a minor gcc release<br>
> ><br>
> > -BINV = "4.9.1"<br>
> > +BINV = "4.9.2"<br>
> ><br>
> > FILESEXTRAPATHS =. "${FILE_DIRNAME}/gcc-4.9:"<br>
> ><br>
> > @@ -69,11 +69,11 @@ SRC_URI = "\<br>
> > file://0051-eabispe.patch \<br>
> ><br>
> file://0053-gcc-fix-segfault-from-calling-free-on-non-malloc-d-a.patch \<br>
> > file://0054-gcc-Makefile.in-fix-parallel-building-failure.patch \<br>
> > - file://0055-PR-rtl-optimization-61801.patch \<br>
> > file://0056-top-level-reorder_gcc-bug-61144.patch \<br>
> > - file://0057-aarch64-config.patch \<br>
> > file://0058-gcc-r212171.patch \<br>
> > file://0059-gcc-PR-rtl-optimization-63348.patch \<br>
> > +<br>
> file://0060-Only-allow-e500-double-in-SPE_SIMD_REGNO_P-registers.patch \<br>
> > + file://0061-Fix-for-unwinder-aborts-on-e500.patch \<br>
> > "<br>
> > SRC_URI[md5sum] = "fddf71348546af523353bd43d34919c1"<br>
> > SRC_URI[sha256sum] =<br>
> "d334781a124ada6f38e63b545e2a3b8c2183049515a1abab6d513f109f1d717e"<br>
> > @@ -125,8 +125,8 @@ EXTRA_OECONF_INTERMEDIATE = "\<br>
> > EXTRA_OECONF_append_libc-uclibc = " --disable-decimal-float "<br>
> ><br>
> > EXTRA_OECONF_PATHS = "\<br>
> > - --with-gxx-include-dir=/not/exist{target_includedir}/c++/${BINV} \<br>
> > - --with-sysroot=/not/exist \<br>
> > +<br>
> --with-gxx-include-dir=${STAGING_DIR_TARGET}${target_includedir}/c++/${BIN<br>
> V} \<br>
> > + --with-sysroot=${STAGING_DIR_TARGET} \<br>
> > --with-build-sysroot=${STAGING_DIR_TARGET} \<br>
> > "<br>
> ><br>
> > diff --git<br>
> > a/meta/recipes-devtools/gcc/gcc-4.9/0060-Only-allow-e500-double-in-SPE<br>
> > _SIMD_REGNO_P-registers.patch<br>
> > b/meta/recipes-devtools/gcc/gcc-4.9/0060-Only-allow-e500-double-in-SPE<br>
> > _SIMD_REGNO_P-registers.patch<br>
> > new file mode 100644<br>
> > index 0000000..810ff90<br>
> > --- /dev/null<br>
> > +++ b/meta/recipes-devtools/gcc/gcc-4.9/0060-Only-allow-e500-double-in<br>
</div></div>> > +++ -SPE_SIMD_REGNO_P-registers.patch<br>
<span class="">> > @@ -0,0 +1,13 @@<br>
> > +diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c<br>
> > +index 512e7d6..292fb2c 100644<br>
> > +--- a/gcc/config/rs6000/rs6000.c<br>
> > ++++ b/gcc/config/rs6000/rs6000.c<br>
> > +@@ -1704,7 +1704,7 @@ rs6000_hard_regno_nregs_internal (int regno, enum<br>
> machine_mode mode)<br>
> > + SCmode so as to pass the value correctly in a pair of<br>
> > + registers. */<br>
> > + else if (TARGET_E500_DOUBLE && FLOAT_MODE_P (mode) && mode != SCmode<br>
> > +- && !DECIMAL_FLOAT_MODE_P (mode))<br>
> > ++ && !DECIMAL_FLOAT_MODE_P (mode) && SPE_SIMD_REGNO_P<br>
</span>> > ++ (regno))<br>
<span class="">> > + reg_size = UNITS_PER_FP_WORD;<br>
> > +<br>
> > + else<br>
> > diff --git<br>
> > a/meta/recipes-devtools/gcc/gcc-4.9/0061-Fix-for-unwinder-aborts-on-e5<br>
> > 00.patch<br>
> > b/meta/recipes-devtools/gcc/gcc-4.9/0061-Fix-for-unwinder-aborts-on-e5<br>
> > 00.patch<br>
> > new file mode 100644<br>
> > index 0000000..c37a4c3<br>
> > --- /dev/null<br>
> > +++ b/meta/recipes-devtools/gcc/gcc-4.9/0061-Fix-for-unwinder-aborts-o<br>
</span>> > +++ n-e500.patch<br>
> > @@ -0,0 +1,161 @@<br>
> > +diff --git a/gcc/defaults.h b/gcc/defaults.h index f94ae17..80a798f<br>
> > +100644<br>
<span class="">> > +--- a/gcc/defaults.h<br>
> > ++++ b/gcc/defaults.h<br>
> > +@@ -438,6 +438,11 @@ see the files COPYING3 and COPYING.RUNTIME<br>
</span>> > +respectively. If not, see #define DWARF_FRAME_REGNUM(REG)<br>
> > +DBX_REGISTER_NUMBER (REG) #endif<br>
<span class="">> > +<br>
> > ++/* The mapping from dwarf CFA reg number to internal dwarf reg<br>
</span>> > ++numbers. */ #ifndef DWARF_REG_TO_UNWIND_COLUMN #define<br>
> > ++DWARF_REG_TO_UNWIND_COLUMN(REGNO) (REGNO) #endif<br>
<span class="">> > ++<br>
> > + /* Map register numbers held in the call frame info that gcc has<br>
> > + collected using DWARF_FRAME_REGNUM to those that should be output in<br>
> > + .debug_frame and .eh_frame. */<br>
</span>> > +diff --git a/gcc/dwarf2cfi.c b/gcc/dwarf2cfi.c index abcdeb3..4e59dfe<br>
> > +100644<br>
<span class="">> > +--- a/gcc/dwarf2cfi.c<br>
> > ++++ b/gcc/dwarf2cfi.c<br>
> > +@@ -252,7 +252,60 @@ init_return_column_size (enum machine_mode mode, rtx<br>
> mem, unsigned int c)<br>
> > + gen_int_mode (size, mode));<br>
> > + }<br>
> > +<br>
> > +-/* Generate code to initialize the register size table. */<br>
> > ++/* Datastructure used by expand_builtin_init_dwarf_reg_sizes and<br>
> > ++ init_one_dwarf_reg_size to communicate on what has been done by the<br>
> > ++ latter. */<br>
> > ++<br>
> > ++typedef struct<br>
> > ++{<br>
> > ++ /* Whether the dwarf return column was initialized. */<br>
> > ++ bool wrote_return_column;<br>
> > ++<br>
> > ++ /* For each hard register REGNO, whether init_one_dwarf_reg_size<br>
</span>> > ++ was given REGNO to process already. */ bool processed_regno<br>
> > ++ [FIRST_PSEUDO_REGISTER];<br>
<span class="">> > ++<br>
> > ++} init_one_dwarf_reg_state;<br>
> > ++<br>
> > ++/* Helper for expand_builtin_init_dwarf_reg_sizes. Generate code to<br>
> > ++ initialize the dwarf register size table entry corresponding to register<br>
> > ++ REGNO in REGMODE. TABLE is the table base address, SLOTMODE is the mode<br>
> to<br>
> > ++ use for the size entry to initialize, and INIT_STATE is the communication<br>
> > ++ datastructure conveying what we're doing to our caller. */<br>
> > ++<br>
> > ++static<br>
> > ++void init_one_dwarf_reg_size (int regno, machine_mode regmode,<br>
> > ++ rtx table, machine_mode slotmode,<br>
> > ++ init_one_dwarf_reg_state *init_state) {<br>
</span>> > ++ const unsigned int dnum = DWARF_FRAME_REGNUM (regno);<br>
<span class="">> > ++ const unsigned int rnum = DWARF2_FRAME_REG_OUT (dnum, 1);<br>
> > ++ const unsigned int dcol = DWARF_REG_TO_UNWIND_COLUMN (rnum);<br>
> > ++<br>
> > ++ const HOST_WIDE_INT slotoffset = dcol * GET_MODE_SIZE (slotmode);<br>
> > ++ const HOST_WIDE_INT regsize = GET_MODE_SIZE (regmode);<br>
> > ++<br>
> > ++ init_state->processed_regno[regno] = true;<br>
> > ++<br>
> > ++ if (rnum >= DWARF_FRAME_REGISTERS)<br>
> > ++ return;<br>
> > ++<br>
> > ++ if (dnum == DWARF_FRAME_RETURN_COLUMN)<br>
> > ++ {<br>
> > ++ if (regmode == VOIDmode)<br>
> > ++ return;<br>
> > ++ init_state->wrote_return_column = true;<br>
> > ++ }<br>
> > ++<br>
> > ++ if (slotoffset < 0)<br>
> > ++ return;<br>
> > ++<br>
> > ++ emit_move_insn (adjust_address (table, slotmode, slotoffset),<br>
> > ++ gen_int_mode (regsize, slotmode)); }<br>
> > ++<br>
</span><span class="">> > ++/* Generate code to initialize the dwarf register size table located<br>
> > ++ at the provided ADDRESS. */<br>
> > +<br>
> > + void<br>
</span>> > + expand_builtin_init_dwarf_reg_sizes (tree address) @@ -261,37<br>
> > ++314,41 @@ expand_builtin_init_dwarf_reg_sizes (tree address)<br>
<span class="">> > + enum machine_mode mode = TYPE_MODE (char_type_node);<br>
> > + rtx addr = expand_normal (address);<br>
> > + rtx mem = gen_rtx_MEM (BLKmode, addr);<br>
> > +- bool wrote_return_column = false;<br>
> > ++<br>
</span>> > ++ init_one_dwarf_reg_state init_state; memset ((char *)&init_state,<br>
> > ++ 0, sizeof (init_state));<br>
<div><div class="h5">> > +<br>
> > + for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)<br>
> > + {<br>
> > +- unsigned int dnum = DWARF_FRAME_REGNUM (i);<br>
> > +- unsigned int rnum = DWARF2_FRAME_REG_OUT (dnum, 1);<br>
> > ++ machine_mode save_mode;<br>
> > ++ rtx span;<br>
> > +<br>
> > +- if (rnum < DWARF_FRAME_REGISTERS)<br>
> > +- {<br>
> > +- HOST_WIDE_INT offset = rnum * GET_MODE_SIZE (mode);<br>
> > +- enum machine_mode save_mode = reg_raw_mode[i];<br>
> > +- HOST_WIDE_INT size;<br>
> > ++ /* No point in processing a register multiple times. This could<br>
> happen<br>
> > ++ with register spans, e.g. when a reg is first processed as a piece<br>
> of<br>
> > ++ a span, then as a register on its own later on. */<br>
> > +<br>
> > +- if (HARD_REGNO_CALL_PART_CLOBBERED (i, save_mode))<br>
> > +- save_mode = choose_hard_reg_mode (i, 1, true);<br>
> > +- if (dnum == DWARF_FRAME_RETURN_COLUMN)<br>
> > +- {<br>
> > +- if (save_mode == VOIDmode)<br>
> > +- continue;<br>
> > +- wrote_return_column = true;<br>
> > +- }<br>
> > +- size = GET_MODE_SIZE (save_mode);<br>
> > +- if (offset < 0)<br>
> > +- continue;<br>
> > ++ if (init_state.processed_regno[i])<br>
> > ++ continue;<br>
> > +<br>
> > +- emit_move_insn (adjust_address (mem, mode, offset),<br>
> > +- gen_int_mode (size, mode));<br>
> > ++ save_mode = reg_raw_mode[i];<br>
> > ++ if (HARD_REGNO_CALL_PART_CLOBBERED (i, save_mode))<br>
> > ++ save_mode = choose_hard_reg_mode (i, 1, true);<br>
> > ++<br>
> > ++ span = targetm.dwarf_register_span (gen_rtx_REG (save_mode, i));<br>
> > ++ if (!span)<br>
> > ++ init_one_dwarf_reg_size (i, save_mode, mem, mode, &init_state);<br>
> > ++ else<br>
> > ++ {<br>
> > ++ for (int si = 0; si < XVECLEN (span, 0); si++)<br>
> > ++ {<br>
> > ++ rtx reg = XVECEXP (span, 0, si);<br>
> > ++ init_one_dwarf_reg_size<br>
> > ++ (REGNO (reg), GET_MODE (reg), mem, mode, &init_state);<br>
> > ++ }<br>
> > + }<br>
> > + }<br>
> > +<br>
> > +- if (!wrote_return_column)<br>
> > ++ if (!init_state.wrote_return_column)<br>
> > + init_return_column_size (mode, mem, DWARF_FRAME_RETURN_COLUMN);<br>
> > +<br>
</div></div>> > + #ifdef DWARF_ALT_FRAME_RETURN_COLUMN diff --git<br>
> > +a/libgcc/unwind-dw2.c b/libgcc/unwind-dw2.c index 55fc4bc..37f0ae2<br>
> > +100644<br>
<span class="">> > +--- a/libgcc/unwind-dw2.c<br>
> > ++++ b/libgcc/unwind-dw2.c<br>
> > +@@ -55,10 +55,6 @@<br>
</span>> > + #define PRE_GCC3_DWARF_FRAME_REGISTERS DWARF_FRAME_REGISTERS #endif<br>
> > +<br>
> > +-#ifndef DWARF_REG_TO_UNWIND_COLUMN<br>
> > +-#define DWARF_REG_TO_UNWIND_COLUMN(REGNO) (REGNO) -#endif<br>
<div class="HOEnZb"><div class="h5">> > +-<br>
> > + /* ??? For the public function interfaces, we tend to gcc_assert that the<br>
> > + column numbers are in range. For the dwarf2 unwind info this does happen,<br>
> > + although so far in a case that doesn't actually matter.<br>
> ><br>
--<br>
_______________________________________________<br>
Openembedded-core mailing list<br>
<a href="mailto:Openembedded-core@lists.openembedded.org">Openembedded-core@lists.openembedded.org</a><br>
<a href="http://lists.openembedded.org/mailman/listinfo/openembedded-core" target="_blank">http://lists.openembedded.org/mailman/listinfo/openembedded-core</a><br>
</div></div></blockquote></div><br></div>